Data analysis and intelligent prediction on X-Ray Diffraction Data using machine learning algorithms under @Professor Shikhar Jha, IITK
Project members : @Pushpanshu Tripathi @Darshit Trevadia
https://github.com/pushpanshu0501/Machine-Learning-and-Data-Analysis-for-Peak-Fitting
(private - for the time being -> contains sensitive data)
All models are trained for data for YSZ (Yttria-stabilized zirconia) & ZnO (Zinc Oxide).
Project deals with prediction of peaks present in X-Ray Diffraction Curves using regression and neural networks. With the data obtained from synchrotron lab in form of Thetas, intensities and image plots, prediction of new plots for similar elements are to be done.
For regression different models were tried and tested finally resulting in Gaussian-Lorentzian regression to be used for peak prediction based upon previous data. Neural networks were used for the same purpose later on.
Using regression, curves are generated outperforming the conventional peak fitting methods by 3 times. Later the generated curves were used for loss calculation from peak medians obtained from CNN model trained on image plots of same datasets. CNN model was trained on calculation of median of image data for all plots and predicting the medians of new plots. Training is done by minimizing the distance from the median from image and median from the plot generated.